Figure 5 4 eth1 active taking over responsibilities for ETH0 This bonding mode

Figure 5 4 eth1 active taking over responsibilities

This preview shows page 74 - 76 out of 142 pages.

Figure 5-4. eth1 active, taking over responsibilities for ETH0 This bonding mode is supported for both Linux bridge and Open vSwitch based bonds and assumes stacked switches ( Example 5-2 ). Example 5-2. Create active-backup bond # xe bond-create mode=active-backup network-uuid=<network-uuid> \ pif-uuids=<pif-uuids> balance-slb bonding With balance-slb bonding, a source-load-balancing active bond is created with bal‐ anced traffic. This means that all interfaces in the bond are used, with traffic balanced between the interfaces based on load and source VM MAC address. Balancing based on source media access control (MAC) address implies that transmit and receive traf‐ fic for a given VM will always occur over the same path. Example 5-3 shows the com‐ mand to create a balance-slb bond. The traffic flow is rebalanced every 10 seconds by default, but this can be defined when the bond is created. When traffic is rebalanced, the physical route to a given VM through the network bond may change. balance-slb uses the standard Linux balance-alb algorithm at its core, which requires the network driver support dynamic rewrite of the MAC address in order to avoid Address Resolution Protocol (ARP)/GARP during rebalancing. This bonding mode is supported for both Linux bridge and Open Virtual Switch based bonds and assumes stacked switches. Note that if network monitoring is done at the physical interface layer, data can be lost during a balancing event. Example 5-3. Create balance-slb bond # xe bond-create mode=balance-slb network-uuid=<network-uuid> \ pif-uuids=<pif-uuids> 60 | Chapter 5: Deployment Blueprint
Image of page 74
LACP bonds LACP bonds are created to support link aggregation or 802.3ad (802.1AX-2008, as defined by the IEEE at ). LACP is only supported by the Open Virtual Switch and requires physical switches configured for LACP using LACP data units (LACPDU). Cisco EtherChannel and Port Aggregation Protocol (PAgP) are not supported. LACP can be configured for up to four interfaces per bond. LACP determines traffic flow by default based on a hash of the source and destination IP, TCP/UDP port, and source MAC address. Example 5-4 shows the command to create a LACP bond. Due to the increased parameters in the hash when compared to balance-slb , LACP bonds can result in transmit and receive traffic to a given VM following different paths. Because destination IP and port are used in the hash, LACP bonds may pro‐ vide limited throughput gains when traffic is occurring to a single destination. An example of such a configuration would be use of LACP for storage traffic where the storage solution is an NFS server, or an iSCSI controller with a single path. In both configurations, the hash will compute to bind traffic to a single member NIC within the bond, which will result in LACP behaving as an active-backup bond.
Image of page 75
Image of page 76

You've reached the end of your free preview.

Want to read all 142 pages?

  • Fall '19
  • Xen, XenServer Dundee

  • Left Quote Icon

    Student Picture

  • Left Quote Icon

    Student Picture

  • Left Quote Icon

    Student Picture